Grounding checks
Every claim must cite an approved passage.
Roadmap
RoadmapGovern · Assurance & Eval Lab
Checks that run before anything reaches a learner
Every release candidate runs a suite of checks: grounding, contradictions, objective coverage, item quality, reading level, accessibility and your own policy rules. Failures block release; warnings need a reviewer's note. Today's drafts already include review notes that flag gaps and claims to verify.
Illustrative preview of the roadmap workbench · fictional sample data
RoadmapA product direction we are designing toward. Not available today. The workbench shows an interactive preview with illustrative data.
Capabilities
What Assurance & Eval Lab does
Contradiction checks
No claim may conflict with a newer source.
Roadmap
Item quality
Objective coverage, plausible distractors, no 'all of the above'.
Roadmap
Readability & accessibility
Reading level targets, alt text, captions, contrast and keyboard access.
Roadmap
Your policy rules
Brand terms, banned claims, required disclaimers.
Roadmap
Review notes today
Every live draft lists source gaps and claims an expert should check.
Live today
How it works
From source to release
STEP 01
Define gates
Pick the checks that must pass and who signs off.
STEP 02
Run on every candidate
Checks run automatically on each Training PR and release.
STEP 03
Fix or justify
Failures block; warnings need a reviewer's note.
STEP 04
Keep the evidence
Results are stored with the release for audits.
Try it
Interactive preview
A front-end demonstration with fictional data, so you can feel the workflow before it ships. Nothing here touches a real system.
Grounding: claims cite approved passagesqueued
Contradictions with newer sourcesqueued
Objective coverage ≥ 2 itemsqueued
Distractor qualityqueued
Reading level ≤ grade 9queued
Inclusive languagequeued
Locale parity (fr-FR)queued
7 checks queued
Depth
Designed for the hard cases
The details that separate a governed system from a generator. All of these are roadmap design, shown so you can judge the direction.
Under the hoodRoadmap
Regression suite
Re-run past checks on every change so fixed problems stay fixed.
Under the hoodRoadmap
Model change qualification
Replay an eval set before adopting a new AI model version, then promote it deliberately.
Under the hoodRoadmap
Critical coverage
Every critical obligation must be taught and assessed.
Under the hoodRoadmap
Policy rules engine
Your own rules as code: required disclaimers, banned claims, terms.
Under the hoodRoadmap
Accessibility critic
Alt text, captions, reading order and contrast checked per release.
Under the hoodRoadmap
Assurance package
The checks, results and sign-offs bundled with each release.
Honest status
What works today, and what is coming
Today
Live today- Review notes flag gaps and claims to check in every draft
- Human review before publishing is built into the workflow
On the roadmap
Roadmap- Automated grounding, contradiction and coverage checks
- Readability and accessibility checks
- Custom policy rules
- Eval suites and tutor red-team runs
FAQ
Questions
Will checks replace expert review?
No. They make review faster and more consistent. A subject-matter expert still approves what ships.
Can we add our own rules?
That's the design: brand, legal and policy rules you define, run on every release.
Keep exploring
Related
Change Intelligence
Most training goes stale quietly. Change Intelligence watches the sources your courses depend on, works out which…
Learn more →
Release & Deploy
Today you export a reviewed course to Word, Markdown, narration scripts or JSON. The roadmap adds real releases:…
Learn more →
Evidence Graph
An evidence graph connects each source clause to the objectives, lessons, questions, scenarios and tutor answers that…
Learn more →
Start with the part that works today
Draft a course from your own document in minutes. The rest of the platform is being built in the open.
Free during early access · No account needed · We don't keep your files